arXiv:adap-org/9909003v1 1 Sep 1999

Analysis of the optimality principles responsible

for vascular network architectonics
I.A.Lubashevsky, V.V.Gafiychuk
Institute of General Physics Academy
of Science of Russian Federation;
Moscow State University, Moscow
Institute of Applied Problems of Mechanics and Mathematics
National Academy of Sciences of Ukraine, Lviv
February 13, 2013
Abstract
The equivalence of two optimality principles leading to Murrays law
has been discussed. The first approach is based on minimization of biological work needed for maintaining the blood flow through the vessels at
required level. The second one is the principle of minimal drag and lumen
volume. Characteristic features of these principles are considered.
An alternative approach leading to Murrays law has been proposed.
For that we model the microcirculatory bed in terms of delivering vascular
network with symmetrical bifurcation nodes, embedded uniformly into
the cellular tissue. It was shown that Murrays law can be regarded as a
direct consequence of the organism capacity for controlling the blood flow
redistribution over the microcirculatory beds.

Introduction

The great amount of natural systems have highly branching networks. As a

evident example of such systems we may regard living tissue where blood supplies the cellular tissue with oxygen, nutritious products, etc. through branching vascular network and at the same time withdraws products resulting from
living activities of the cellular tissue. A similar situation takes place in respiratory systems where oxygen reaches small vessels (capillaries) going through
the hierarchical system of bronchial tubes. There is a question of what physical
principles govern network organization living systems under consideration. In
this paper we focus our attention on the analyzes of different known optimal
principles of network formation for microcirculatory bed and developing new
approach to this problem.

Table 1: Typical parameters of the vessel arrangement of a 13-kg dog, after [30].
Diameter Length Tissue/Vessel Arrangement
Vessel type
2a, m
l, cm
radii, d/a
anisotropy
Primary arteries
Small arteries
Terminal vessels
Arterioles

300
100
50
20

1.0
0.5
0.2
0.1

30
20
10
7

0.86
0.50
0.37
0.25

measured as (d2 l)1/3 /l

Analysis of the optimality principles

A microcirculatory bed can be reasonably regarded as a space-filling fractal

being a natural structure for ensuring that all cells are serviced by capillaries
[31]. The vessel network must branch so that every small group of cells, referred
below to as elementary tissue domain, is supplied by at least one capillary.
Since a typical length of capillaries is about 0.3 to 0.7 mm a vessel network
generated by an artery of length of order of 1 to 5 cm should contain a sufficiently
large number of hierarchy levels. At zeroth level we meet the host artery and the
host vein, the mother and daughter vessels belong to n-th and (n + 1)-th levels,
respectively, and the last level N comprises capillaries. So at each level n of the
vascular network the tissue domain supplied by a given microcirculatory bed as
a whole can be approximated by the union of the tissue subdomains whose mean
size is about the typical length ln of the n-th level vessels. Thus, the individual
volume of these subdomains is estimated as Vn ln3 and their total number (as
well as the total number of n-th level vessels) is about Mn V0 /ln3 , where V0 is
the total volume of the microcirculatory bed. The higher is the level, the more
accurate become the independence of such estimates from the particular details
of vessel arrangements. For internal organs they approximately hold also for
large vessels of regional circulation. To justify the latter statement we present
Table 1 relating the vessel lengths and radii to the radii of the corresponding
tissue cylinders, i.e. the cylindrical neighborhood falling on one vessel of a fixed
level.
This condition that the vascular network be volume-preserving from one
generation to the next gives us immediately the local relation between the char3
acteristic lengths of the vessels: ln3 gln+1
(here g = 2 is the order of the
vessel branching node). Whence it follows that ln l0 g n/3 , where l0 is the
characteristic size of the microcirculatory bed region or, what is practically the
same, the length of the host artery.
The following analysis, however, will require a more detailed information
about the vascular network architectonics. Namely, we need to know how the
vessel radii change at the nodes and the relative arrangement of mother and
daughter vessels. Actually here we meet the problem as to what fundamental

2
>

Figure 1: Characteristics of the vessel branching

regularities govern the vessel branching. These regularities manifest themselves
in the relation between the radii a0 , a1 , a2 of mother and daughter arteries,
respectively, and the angles 1 , 2 , 12 = 1 + 2 which the daughter branches
make with the direction of the mother artery and with each other (Fig. 1).
A detailed theoretical attempt of understanding this regularity was made
first by Cecil D. Murray in 1926 [19, 20]. He proposed a model relating the
artery radii at branching nodes (Fig. 1) by the expression
ax0 = ax1 + ax2

with x = 3

(1)

thereafter referred to as Murrays law (x is also called the bifurcation exponent).

Then Murrays approach was under development in a large number of works, see,
for example, [1, 2, 22, 29] and a series of works by Zamir et all. [36, 37, 40, 41]
and by Woldenberg et all. [33] (for a historical review see also [28, 40, 12]).
The idea of Murrays model is reduced to the assumption that physiological
vascular network, subject through evolution to natural selection, must have
achieved an optimum arrangement corresponding to the least possible biological
work needed for maintaining the blood flow through it at required level. This
biological work P involves two terms: (i) the cost of overcoming viscous drag
during blood motion through the vessels obeying Poiseuilles law, and (ii) the
energy metabolically required to maintain the volume of blood and the vessel
tissue. Dealing with an individual artery of length l and radius a with a blood
flow rate J in it we get:
P=

8lJ 2
+ ma2 l ,
a4

(2)

where is the blood viscosity and m is a metabolic coefficient. Minimizing

function (1) with respect to a we find the relation between the blood flow rate J
and the artery lumen radius a corresponding to the given optimality principles:
J = ka3 ,
3

(3)

Table 2: Integral characteristics of the vessel cross-section at different branching

levels, after [12].
P 2
P 3
P 4
Mean radius
a
a 10
a
Vessel type
a, m
cm2
cm3
cm4
Homo sapiens

Aorta
Arteries
Arterioles
Capillaries
Venules
Veins
Vena cava

12500
2000
30
6
20
2500
15000

1.56
6.36
127.4
1432
1273
12.9
2.25

1.95
1.27
0.382
0.860
2.55
3.18
3.38

Canis familiaris

Aorta
Large arteries
Main arterial
branches
Terminal branches
Arteriolas
Capillaries
Venules
Terminal veins
Main venous
branches
Large veins
Vena cava

5000
1500
500

1.25
1.35
0.75

300
10
4
15
750
1200

0.486
0.400
0.768
2.70
7.59
10.37

3000
6250

10.80
1.53

2.44
0.25
1.15 103
5.16 104
5.09 103
0.80
5.06

p
where the coefficient k = m 2 /(16) is a constant for the tissue under consideration. Due to the blood conservation at branching nodes we can write
J0 = J1 + J2 (Fig. 1) whence Murrays law (1) immediately follows.
However, care must be taken in comparing measurements with prediction,
particularly if averages over many successive levels are used. Already Mall
himself noted that his data were approximate [17]. In particular, for large
arteries of systemic circulation where blood flow can be turbulent the bifurcation
exponent x should be equal to 7/3 2.33 as it follows from this optimality
principle of minimum pumping power and lumen volume.
There is also another optimality principle leading to Murrays law, the principle of minimum drag and lumen surface [36, 37]. The drag against the blood
motion through vessels is caused by the blood viscosity and can be described in
terms of the shear stress on the walls of vessels, 2alt , where
t = n v =

4 J
a
P =
2l
a3

(4)

for the laminar flow and P is the pressure drop along a vessel of length l
and radius a. Then the given optimality principle is reduced to the minimum
condition for the function
P =

8lJ
+ m 2al ,
a2

(5)

where m is a certain weighting coefficient. Minimizing (5) with respect to a we

get a relationship between J and a of the same form as (3), leading to Murrays
law again. There were a number works (see, e.g., [40, 33, 5, 7]) aimed at finding out what the specific optimality principle governs the artery branching by
studying the angles of daughter vessels, 1 , 2 , 12 , in relation to the asymmetry
of the branching node, a2 /a1 (Fig. 1). However, on one hand, all the optimality principles give numerically close relationships between the vessel angles and
radii for the bifurcation exponent x 3 [33]. On the other hand, it turned out
that experimentally determined branching angles generally exhibit considerable
scatter around the theoretical optimum. The matter is that small variations of
the total cost of artery bifurcation about several percents causes the actual
vessel angles to deviate significantly from the predicted optimum. This feature
is illustrated in Fig. 2 showing the variations in the vessel angles governed by the
minimality of functional (5) with imposed 10% perturbations. Namely, varying
the coordinates of the branching node (Fig. 1) we get that the minimum of the
function P0 + P1 + P2 is attained when
a1 cos 1 + a2 cos 2

(1 + )a0 ,

(6)

a1 sin 1 a2 sin 2

a0 ,

(7)

where the additional terms a0 and a0 with || , | | < 0.1 describes possible
deviations from the optimality condition. The resulting values of 1 and 2
are depicted in Fig. 2. It should be noted that expressions (6), (7) correspond
actually to the mechanical equilibrium of the node under the action of vessel
5

180

1. mother - smaller daughter

2. mother - larger daughter

Vessel angle (degrees)

150
120

90
60
30

2
0
0.0

0.2

0.4

0.6

0.8

1.0

Asymmetry ratio,
Figure 2: The angles that the daughter arteries make with the mother artery
when the vessel branching is governed by the minimum drag & suface principle
(solid lines) and under 10% perturbations (darkened region).
walls strained by the blood motion and the additional terms describe a possible
effect of the cellular tissue.
Nevertheless, the optimality principles based on functional (2) seems to govern the artery bifurcations [7]. Besides, this principles gives also adequate estimates of the integral characteristics of microcirculatory beds [10, 11].
The bifurcation exponent x, on the contrary, is well approximated by the
Murray value, x 3, at least starting from arteries of intermediate size [39].
This value meets also the space-filling requirement for the vascular network
fractal in geometry to fill precisely the space of a fixed relative volume at each
hierarchy level [18]. Indeed, assuming the volume of the tissue cylinders matching an artery of length l and lumen radius a to be about l3 we get that the
corresponding relative volume of blood is (a/l)2 . So it is fixed if a = constant l
and, thus, a30 = a31 + a32 provided the tissue cylinder matching the mother artery
is composed of the tissue cylinders of the daughter arteries.
In order to specify the microcirculatory bed structure we need also to classify vessels according to the symmetry of their branching (Fig. 3). The matter
is that [41] arteries with predominantly asymmetric bifurcations give off comparatively little flow into its side branches along its way and, therefore, able
to carry the mainstream flow across larger distances. Conversely, a more symmetric bifurcation pattern splits flow into numerous small branches, thereby
delivering blood to its surrounding tissue. Such arteries have been attributed a
conveying and delivering types of function, respectively. Since blood must
be conveyed towards the sites at which to be delivered, both types of vessels
occur in real arterial trees. Moreover, a larger conveying vessel may switch into

3
2
1
0
a

levels

Figure 3: Schematic illustration of the conveying (a) and deliverying (b)

types of artery pattern.
a bunch of small delivering branches.
Obviously, real arterial trees should contain a great variety of intermediate
stages in between these extremes and as our field of view moves from the large
systemic arteries to small arteries of regional circulation the vessel bifurcation
should become more and more symmetrical. This has been also justified by numerically modelling the structure of arterial trees governed by the minimality
condition of blood volume [24, 25, 26, 27]. According to the experimental data
(see, e.g., the work [5] and Fig. 4 based on it) even sufficiently large regional
arteries of diameter and length about 300 m and 1 cm, respectively, (Table 1)
branch symmetrically, at least at first approximation.. Therefore microcirculatory beds as they have been specified above can be regarded as a vessel network
with approximately symmetrical bifurcations.
In other words, we may think of the systemic arteries as vessels of the conveying type where the mean blood pressure is practically constant. Conversely,
the arteries of microcirculatory beds should belong to the delivering type and
mainly determine the total resistance of the vascular network to blood flow,
with the blood pressure drop being uniformly distributed over many arteries of
different length.

Physiological mechanisms governing the vessel arrangements

The universality of Murrays law for distributed transport systems in many different live organisms raises questions as to: What cues are available to organisms
to use in generating such systems? What physiological mechanisms enable them
to adapt to altering conditions? Do in fact live organisms follow certain global
optimality principles?
For Murrays law (3) the shear stress t is constant (see formula (4)) throughout a given artery system. Rodbard [21] proposed that the shear stress detected
by the vessel endothelium leads to the vessel growth or contraction, and Zamir
[38] suggested that this leads to the development of Murrays system as vessels
7

branching symmetry,

1.00

0.75

0.50

0.25

0.00
10

100

1000

parent diameter, 2

10 000

(m)

Figure 4: Bifurcation symmetry vs. vessel diameter for the porcine coronary
arterial tree (based on the data of [5].) The darkness intensity indicates the
density of the experimental points.
maintain a constant value of shear stress. Concerning the particular mechanism by which organisms can implement the shear stress sensitivity we can say
the following. Now it is established that the adaptation of conduit arteries as
well as resistance arteries to acute changes in flow is mediated by the potent
endogenous nitrovasodilator endothelium-derived relaxing factor, whose release
from endothelial cells is enhanced by flow through the physical stimulation of
shear stress (see, e.g., [7] and references therein). The adaptation of arterial diameters to long-term changes in the flow rate also occurs through a mechanism
which appears to involve the sensitivity to shear stress and the participation of
endothelial cells, but remains not to be understood well [7].
It should be noted that the shear stress equality through a vascular network
does not lead directly to a certain optimality principle. Different principles,
for instance, (2) and (5), can give the same condition imposed on the shear
stress. Moreover, it is quite possible that the case of this equality is of another
nature. In particular, for large conduit arteries in the human pulmonary tree
the bifurcation exponent x is reported to be in the range 12, whereas Murrays law holds well beginning from intermediate conveying arteries [34, 13].
The matter is that in large systemic arteries the blood pressure exhibits substantial oscillations because of the heart beating, giving rise to damped waves
travelling through the systemic arteries. The value of the bifurcation exponent
x = 2 matching the area-preserving law at the branching nodes ensures that the
energy-carrying waves are not reflected back up the vessels at the nodes. How-

ever, this requirement is also can be derived from a certain optimality principle
[31].
Summarizing the aforesaid we will model the microcirculatory bed in terms
of a delivering vascular network with symmetrical bifurcation nodes embedded
uniformly into the cellular tissue. Besides, the Murrays law will be assumed to
hold. The latter is also essential from the standpoint of the tissue self-regulation,
which will be discussed in detail in the next section. Here, nevertheless, we make
several remarks concerning the given aspect too, because it could be treated
as an alternative origin of Murrays law (3). Let us consider a symmetrical
dichotomous vessel tree shown, e.g., in Fig. 3b. In order to govern blood flow
redistribution over the microcirculatory bed finely enough so to supply with, for
example, increased amount of blood only those regions where it is necessary and
not to disturb other regions the blood pressure should be uniformly distributed
over the microcirculatory bed, at least, approximately.
The blood pressure drop Pn along an artery of level n (n = 0, 1, 2, . . . ,
Fig. 3) for laminar blood flow is
Pn =

8ln Jn
.
a4n

(8)

For the space-filling vascular network this artery supplies with blood a tissue
region of volume about ln3 and, so, under normal conditions the blood flow rate
Jn in it should be equal to Jn jln3 , where j is the blood perfusion rate (the
volume of blood flowing through a tissue domain of unit volume per unit time)
assumed to be the same at all the points of the given microcirculatory bed.
Then formula (8) gives us the estimate
 4
8j ln
Pn =

an
whence it follows that Pn will be approximately the same for all the levels, i.e.
the blood pressure will be uniformly distributed over the arterial bed if the ratio
ln /an takes a certain fixed value, ln constantan and, thus, Jn constant a3n .
Due to the blood conservation at branching nodes we can write
J0 = J1 + J2

(9)

(see Fig. 1). The later gives us immediately Murrays law (1). In other words,
Murrays law can be also regarded as a direct consequence of the organism capacity for controlling finely the blood flow redistribution over the microcirculatory
beds.
It should be noted that in the previous papers [14, 15, 16] we considered in
detail the mathematical model for the vascular network response to variations
in the tissue temperature on the given network architectonics. We have found
that the distribution of the blood temperature over the venous bed aggregating
the information of the cellular tissue state allows the living tissue to function
properly. We showed that this property is one of the general basic characteristics
of various natural hierarchical systems. These systems differ from each other
by the specific realization of such a synergetic mechanism only.
9

References
[1] Cohn, D. Optimal systems I: the vascular system, Bull. Math. Biophys.,
16, pp. 5974 (1954).
[2] Cohn, D. Optimal systems II: the vascular system, Bull. Math. Biophys.,
17, pp. 219227 (1955).
[3] Chernoushko, F. L. Optimal branching tree structures, Appl. Math.
Mech. U.S.S.R., 41, pp. 376383 (1977).
[4] Chernoushko, F. L. Optimal branching structures in biomechanics, Mech.
Compos. Mater. U.S.S.R., 16, pp. 308313 (1980).
[5] van Bavel, E. and Spann, J. A. E. Branching patterns in the porcine
coronary arterial tree. Estimation of flow heterogeneity, Cir. Res., 71,
pp. 12001212 (1992).
[6] Green, H. D. Circulation: physical principles, in: Medical Physics,
O. Grasser, edit. (Year Book Publishers, Chicago, 1944), p. 210.
[7] Griffith, T.M. and Edwards, D.H. Basal EDRF activity helps to keep
the geometrical configuration of arterial branchings close to the Murray
optimum, J. Theor. Biol., 146, pp. 545573 (1990).
[8] Kamiya, A. and Togawa, T. Optimal branching structure of the vascular
tree, Bull. Math. Biophys., 34, pp. 431438 (1972).
[9] Kamiya, A., Togawa, T., and Yamamota, A. Theoretical relationship between the optimal models of the vascular tree, Bull. Math. Biophys., 34,
pp. 311326 (1974).
[10] Khanin, M.A. and Bukharov, I.B. Optimal arterial blood pressure, J.
Theor. Biol., 148, pp. 289294 (1991).
[11] Khanin, M.A. and Bukharov, I.B. Optimal structure of the microcirculatory bed, J. Theor. Biol., 169, pp. 267273 (1994).
[12] LaBarbera, M. Principles of design of fluid transport systems in zoology,
Science, 249, pp. 9921000 (1990).
[13] Li, K.-J. Comparative Cardiovascular Dynamics of Mammals (CRC Press,
Boca Raton, FL., 1996).
[14] Lubashevsky,
I.A.,
Gafiychuk, V.V. Cooperative mechanism
of self-regulation in hierarchical living systems. e- print.
http://xxx.lanl.gov/abs/adap-org/9808003 (1998) to be published in
SIAM J.Appl.Math.
[15] Lubashevsky, I.A., Gafiychuk, V.V. Bioheat transfer. New approach to diffusive type process in hierarchically organized active heterogenous media.
(Book to be published).
10

[16] Lubashevskii, I.A., Gafiychuk, V.V. A simple model of self-regulation in

large natural hierarchical systems. J.Env. Systems. - v.23, No.3. p.281-289.
(1995).
[17] Mall, F. P. Die blut und lymphwege in dunndarm des hundes, Abhandlungen der Mathematisch-Physischen Classe der Koniglich Sachsischen Gessellschaft der Wissenschaften, 14, pp. 151200 (1888).
[18] Mandelbrot, B.B. The Fractal Geometry of Nature (Freeman, New York,
1982).
[19] Murray, C. D. The physiological principle of minimum work. I. The vascular system and the cost of blood volume, Proc. Natl. Acad. Sci. U.S.A.,
12, pp. 207214 (1926).
[20] Murray, C. D. The physiological principle of minimum work applied to
the angle of branching of arteries, J. Gen. Physiol., 9, pp. 835841 (1926).
[21] Rodbard, S. Vascular caliber, Cardiology, 60, pp. 449 (1975).
[22] Rosen, R. Optimal principles in biology ( Butterworths, London, 1967).
[23] Schleier, J. Der Energieverbrauch in der blutbahn, Pflugers Archiv
Gesamte Physiol. Menschen Tiere, 173, pp. 172204 (1919).
[24] Schreiner, W. Computer generation of complex arterial tree models, J.
Biomed. Eng., 15, pp. 148150 (1993).
[25] Schreiner, W. and Buxbaum, P.F. Computer-optimization of vascular
trees, IEEE Trans. Biomed. Eng., 40, pp. 482491 (1993).
[26] Schreiner, W., Neumann, M., Neumann, F.,and Buxbaum, P.F.
Computer-optimization of vascular trees, IEEE Trans. Biomed. Eng.,
40, pp. 482491 (1993).
[27] Schreiner, W., Neumann, F., Neumann, M., End, A., and M
uller, M.R.
Structural quantification and bifurcation symmetry in arterial tree models
generated by constrained constructive optimization, J. Theor. Biol., 180,
pp. 161174 (1996).
[28] Sherman, T. F. On connecting large vessels to small. The meaning of
Murrays law, J. Gen. Physiol., 78, pp. 431453 (1981).
[29] Shoshenko, K.A., Golub, A.G., Brod,V.I., Barbashina, N.E., Ivanova, S.F.,
Krivoshapkin, A.L., and Osipov V.V. Architectonics of the Circulatory Bed
(Nauka Publishers, Novosibirsk, 1982) [in Russian].
[30] Weinbaum, S., Jiji, L.M., and Lemons,D.E. Theory and experiment for
the effect of vascular microstructure on surface tissue heat transfer - Part
I: Anatomical foundation and model conceptualization. Trans. ASME J.
Biom. Eng., 106, pp. 321330 (1984).
11

[31] West, G. B., Brown, J. H., and Enquist, B. J. A general model for the
origin of allometric scaling laws in biology, Science, 276, pp. 122126
(1997).
[32] Wilde, A. G. The branching patterns of blood-vessels (a theoretical approach), Z. Morphol. und Anthropol., 57, pp. 4155 (1965).
[33] Woldenberg, M.J. and Horsfield, K. Relation of branching angles to optimality for four cost principles, J. Theor. Bilog., 122, pp. 187204 (1986).
[34] Woldenberg, M.J. and Horsfield, K. Finding the optimal lengths for three
branches at a junction, J. Theor. Bilog., 104, pp. 301318 (1983).
[35] Young, T. On the functions of the heart and arteries, Philos. Trans. Royal
Soc. London., pp. 1-31 (1809).
[36] Zamir, M. Optimality principles in arterial branching, J. Theor. Bilog.,
62, pp. 227251 (1976).
[37] Zamir, M. The role of shear forces in arterial branching, J. Gen. Physiol.,
67, pp. 213222 (1976).
[38] Zamir, M. Shear forces and blood vessel radii in the cardiovascular system, J. Gen. Physiol., 69, pp. 449461 (1977).
[39] Zamir, M., Wrigley, S.M., and Langille, B.L. Arterial bifurcations in the
cardiovascular system of a rat, J. Gen. Physiol., 81, pp. 325335 (1983).
[40] Zamir, M. The branching structure of arterial trees, Comm. Theor. Biol.,
91, pp. 1537 (1988).
[41] Zamir, M. Distributing and delivaring vessels of the human heart, J. Gen.
Physiol., 91, pp. 725735 (1988).

12